Southwest Business and Economics Journal/2003-2004
54
Additional Minimum Pension
Liability and Pension Underfunding
Additional Minimum Pension Liability and Pension Underfunding
Barbara W. Scofield
Paul Haensly
The University of Texas of the Permian Basin
Abstract
This study examines two views of additional minimum pension liability (AMPL). The traditional
view emphasizes AMPL’s computational precedents: accumulated benefit obligation, prepaid (accrued)
pension cost, and pension plan assets. The component view emphasizes the underlying unrecognized prior
service cost, unrecognized net losses (gains), estimated rate of pension increase, and accumulated benefit
obligation. This study is an initial exploration of the relationship between AMPL recognition, its computational
precedents, and its underlying components. We find that AMPL recognition is more strongly associated with
its underlying components and thus provides support for the component view of AMPL, paving the way for
further evaluation of AMPL.
Introduction
Additional minimum pension liability (AMPL) is a constructed accounting concept created by the
FASB in SFAS 87. AMPL was designed to protect investors, creditors, and other decision-makers from
underestimating the liabilities of a company with an underfunded pension plan. It occurs when both
underfunding and sufficient off-balance sheet deferrals exist as a result of the company’s use of smoothing
techniques in its GAAP pension accounting. Research in this area has either ignored AMPL and explored
underfunding alone or it has considered AMPL solely as a result of the pension investment deferrals. This
study looks at companies that disclosed AMPL in their 1993-1996 annual reports in order to document the
underlying determinates of AMPL. This preliminary study is important because the usual computation of
AMPL uses summary variables that include components that are open to varied interpretations. Although
preparers use accumulated benefit obligation, plan assets at fair value, and prepaid (accrued) pension cost to
compute AMPL, the underlying components include unrecognized prior service cost, unrecognized net losses
(gains), and estimated rate of compensation increase.
This paper tests the extent to which additional minimum pension liability is more closely associated
with the rate of compensation increase, unrecognized net gains and losses, and unrecognized prior service
cost rather than with underfunding alone. We develop a model that demonstrates that plan assets are relevant
in establishing the existence of additional minimum pension liability but not in its quantitative measurement.
Thus two pension plans that are equally underfunded may have different additional minimum pension liability
because one plan may have few deferrals and more liability already on the books of the company and another
plan may have many deferrals and less liability already on the company’s books.
This paper shows that high levels of unrecognized prior service cost (UPSC), high levels of
unrecognized net losses (gains) (UGL), and low levels of estimated rate of compensation increase (RCI) are
significantly related to AMPL recognition while levels of plan assets at fair value (PA) and prepaid (accrued)
pension cost (PAPC) generally are not. Thus, the traditional formulation of AMPL does not adequately identify
the factors that are most important to AMPL recognition.
The importance of this conclusion is that each of these factors, which are disclosed in financial
statement footnotes, can imply different company positions. For example, UPSC may result from competition
in the marketplace for labor. If decision-makers are satisfied with a company’s workforce, high levels of
UPSC may indicate positive management decisions that decrease employee turnover and maintain a company’s
human resource pool. On the other hand, high levels of unrecognized net losses (low levels of unrecognized
55
Southwest Business and Economics Journal/2003-2004
net gains) may result from macro-economic changes to which a company cannot adjust. Poor investment
performance, unrealistic estimates of returns on plan assets, and changes in actuarial assumptions may increase
UGL beyond what the company can currently fund. Decision-makers could easily be dissatisfied with a
company with an AMPL based on one of these factors. Thus it becomes important to correctly identify those
factors that underlie AMPL recognition.
Literature Review
Additional minimum pension liability has seldom been mentioned in academic accounting research
or in the business media, although pension accounting remains a current issue in accounting standard setting.
Academic research typically evaluates pensions using the entire liability and asset amounts from footnotes.
In fact, empirical work with AMPL had been tagged as unnecessary because AMPL is “both small and relatively
rare” (Hand, Landsman & Monahan, 1998). Literature examining pensions is consistent, however, with AMPL’s
having some importance. Chen and Lin (1992), Kemp (1987), Sharad (1999), and Newell, Kreuze, and Hurtt
(2002) differentiate the behavior of companies with underfunded pension plans from those with funded plans.
Not all underfunded plans, however, exhibit AMPL. Amir and Benartzi (1999) examine underfunded plans to
determine whether firms use their pension asset allocation to avoid recognition of AMPL. The greater use of
fixed investments improves the predictability of investment returns and makes it less likely that unexpected
investment losses will balloon deferrals and give rise to AMPL in the presence of underfunding.
More recent academic work has looked at the income statement effects of required GAAP pension
smoothing techniques. The investment and amortization components of pension expense, and thus net income,
are the result of the the same smoothing techniques that create AMPL on the balance sheet. In the absence of
UGL or UPSC, neither the expense allocations nor AMPL would exist. Coronado and Sharpe (2003) identify
mispricing of corporate stock due to the market’s applying a common P/E ratio to pension earnings and
operating earnings. The articulation of income statements and balance sheets suggests that the effect of
smoothing on balance sheets may be similarly opaque to investors. Additional indirect evidence about the
impact of AMPL is found in Harper and Strawser (1993). They document that firms increased their pension
funding when they adopted SFAS No. 87 because more conservative reporting and greater disclosures are
required with SFAS No. 87. The presence of AMPL, an increase to liabilities, was introduced in SFAS No.
87.
In the business media, AMPL’s occasional appearance suggests significant misunderstandings of
AMPL. The Investment Dealer’s Digest states, “Because companies with underfunded pensions would
ultimately be required to make an additional minimum liability adjustment to their balance sheets, that liability
might trigger debt rating agencies’ concerns” (Tunick 2002, 10). However additional minimum pension liability
may never appear for some underfunded companies because AMPL will only occur in the joint presence of
underfunding and the presence of off-balance sheet deferrals. More often only the income statement impact
of GAAP pension accounting and not the accompanying AMPL consequences are mentioned in the business
press. Widely lamented is the netting of service cost with estimated pension investment income and use of
smoothing procedures to spread pension asset investment gains and losses across future periods, decreasing
the transparency of financial statements (for example, McGough & Schultz, 1999; Byrnes, 2000; Henry,
2001; Dale, 2002; MacDonald, 2003).
GAAP pension expense includes allocated interest and amortizations of deferred amounts. This
benefit or burden to current operating earnings has led to a reaction against the unadjusted use of pension
expense in decision-making. Standard and Poor’s (S&P) has eliminated the investment component of pension
expense from its measure of core earnings (Blitzer, Friedman & Silverblatt, 2000). The Wall Street Journal
reports that many loan agreements exclude pension components of other comprehensive income in evaluating
equity (Bryan-Low, 2002). Kranhold and Schultz (2003) report that General Electric adjusts GAAP income
for the provision for pensions before using an income measurement in computing executive salaries. Bader
(2003) recommends that financial analysts value corporations using only the service cost component of pension
expense and adjusting the balance sheet to include the fair value of pension assets and pension accumulated
benefit obligation. The balance sheet expression of the smoothing techniques, AMPL, is mentioned in
56
Additional Minimum Pension
Liability and Pension Underfunding
connection with S&P’s credit rating function (Bryan-Low, 2002). Thus, there is much current interest in the
distortions of GAAP pension accounting, including direct and indirect interest in AMPL.
The Financial Accounting Standards Board has reexamined the place of AMPL in accounting standardsetting. When the Financial Accounting Standards Board reconsidered disclosure effectiveness in 1995, pension
accounting became an initial target (FASB, 1995). In fact the exposure draft on Employers’ Disclosures about
Pensions and Other Postretirement Benefits (FASB, 1997) includes, “financial analysts commented that
information about unamortized balances of prior service cost and transition amounts is useful in assessing
current earnings and forecasting future amortization (paragraph 3).” Prior service cost is a basic factor that
can lead to additional minimum pension liability. FASB’s initial revision of pension disclosures in Statement
of Financial Accounting Standards No. 132 (SFAS 132) provided clearer disclosures about all pension-related
accounts on the balance sheet and maintained the disclosure of accumulated benefit obligation for underfunded
pension plans, which is used in GAAP accounting only for AMPL computation. Thus there is some indication
that FASB considered AMPL information useful to decision-makers and therefore made some underlying
disclosures more accessible. An alternative to providing copious disclosures about recognized items is to
fundamentally clarify pension accounting so that fewer explanatory disclosures are required. Certainly pension
accounting would be simplified without companies’ reporting AMPL in addition to accruing pension expense.
However, the loss of AMPL information without offsetting changes in the recognition and measurement of
pension elements may be costly to information users. While pension disclosures have been changed again
with Statement of Financial Accounting Standards No. 132 (revised 2003), FASB has not yet added the
recognition and measure of pensions to its agenda.
Expanding the present understanding of the role of additional minimum pension liability in financial
reporting can help financial statement users and standard-setters. This paper begins with a description of
additional minimum pension liability and analyzes the role of its components.
Description of Additional Minimum Pension Liability
AMPL is the excess pension liability not covered by plan assets and not already disclosed in financial
statements. It is calculated from the accumulated benefit obligation (ABO), plan assets at fair value (PA), and
prepaid (accrued) pension cost (PAPC).
Under the traditional formulation, AMPL is determined by a two-stage calculation. First, minimum
pension liability (MPL) is calculated as
MPL = max[ABO – PA, 0]
(1)
If MPL is positive, then AMPL is calculated as
AMPL = max[MPL + PAPC, 0]
(2)
where PAPC is a positive number if it is a prepaid cost, i.e., an asset; PAPC is a negative number if
it is an accrued cost, i.e., a liability. Because AMPL is nonzero only if MPL > 0, the two-stage computation
can be represented as
AMPL = I(MPL>0)·max[ABO – PA + PAPC, 0]
(3)
where I(MPL>0) is a dummy variable that equals one if MPL is positive and zero otherwise.
Defining prepaid (accrued) pension cost in terms of its components suggests an alternative way to view
AMPL. PAPC can be calculated as
PAPC = -PBO + PA + UPSC + UGL
(4)
57
Southwest Business and Economics Journal/2003-2004
where PBO is the projected benefit obligation; PA is plan assets at fair value; UPSC is unrecognized
prior service cost; and UGL is unrecognized net losses (gains). In this formulation, UGL is a positive number
if a net loss and a negative number if a net gain. If PBO is approximated by
PBO ≅ (1 + RCI)ABO
(5)
where RCI is estimated rate of compensation increase, then PAPC may be approximated by
PAPC ≅ -(1 + RCI)ABO + PA + UPSC + UGL
(6)
Substituting this approximation into equation (3) gives the following approximation to AMPL:
AMPL ≅ I(MPL>0)·max[-RCI·ABO + UPSC + UGL, 0]
(7)
where MPL = max[ABO – PA, 0].
The traditional and component formulations yield different predictions on the effects of changes in
pension plan components. The purpose of this paper is to examine which formulation of AMPL has more
empirical support. The traditional formulation in equation (3) indicates that companies are more likely to
have AMPL as ABO increases, PA decreases, and PAPC increases (i.e., becomes a larger asset or smaller
liability). The component approach indicates that companies are more likely to have AMPL as PA decreases,
as RCI decreases, UPSC increases (i.e., is a larger asset or smaller liability), and UGL increases (i.e., is a
larger unrecognized net loss or smaller unrecognized net gain). The effect of increases in ABO depends on
whether ABO is greater or less than PA, so ABO is not helpful in distinguishing between the two views.
Hypotheses
This paper examines five hypotheses concerning the effects of changes in pension plan components
under traditional and component formulations of AMPL. The hypotheses compare firms with underfunded
defined benefit pension plans with AMPL to those without.
Hypothesis 1. Under the traditional formulation, firms with and without AMPL have no significant
difference in estimated rate of compensation increase. Under the component formulation, firms with no AMPL
should have higher levels of RCI than firms with AMPL. The statistical test takes the following form:
H0: mean of RCI (no AMPL) d” mean of RCI (AMPL)
HA: mean of RCI (no AMPL) > mean of RCI (AMPL)
Hypothesis 2. Under the traditional formulation, firms with and without AMPL have no significant
difference in unrecognized prior service cost. Under the component formulation, firms with no AMPL should
have lower levels of UPSC than firms with AMPL. The statistical test takes the following form:
H0: mean of UPSC (no AMPL) e” mean of UPSC (AMPL)
HA: mean of UPSC (no AMPL) < mean of UPSC (AMPL)
Hypothesis 3. Under the traditional formulation, firms with and without AMPL have no significant
difference in unrecognized net losses (gains). Under the component formulation, firms with no AMPL should
have lower levels of UGL (smaller unrecognized net losses or larger unrecognized net gains) than firms with
AMPL. The statistical test takes the following form:
H0: mean of UGL (no AMPL) e” mean of UGL (AMPL)
HA: mean of UGL (no AMPL) < mean of UGL (AMPL)
58
Additional Minimum Pension
Liability and Pension Underfunding
Hypothesis 4. Under both the traditional and component formulations, firms with no AMPL should
have higher levels of pension assets than firms with AMPL. The statistical test takes the following form:
H0: mean of PA (no AMPL) d” mean of PA (AMPL)
HA: mean of PA (no AMPL) > mean of PA (AMPL)
Hypothesis 5. Under the component formulation, firms with and without AMPL have no significant
difference in prepaid (accrued) pension cost. Under the traditional formulation, firms with no AMPL should
have lower levels of PAPC (smaller prepaid or larger accrued pension cost) than firms with AMPL. The
statistical test takes the following form:
H0: mean of PAPC (no AMPL) e” mean of PAPC (AMPL)
HA: mean of PAPC (no AMPL) < mean of PAPC (AMPL)
The two views lead to different predictions for the relationships between AMPL and the pension
components RCI, UPSC, UGL, and PAPC (hypotheses 1, 2, 3, and 5). If the traditional view of AMPL is
more appropriate, then the null hypotheses 1, 2, and 3 will not be rejected, while null hypotheses 4 and 5 will
be rejected in the predicted direction. If the component view of AMPL is more appropriate, then null hypotheses
1, 2, 3, and 4 will be rejected in the predicted direction, while null hypothesis 5 will not be rejected.
In cross-sectional comparisons, levels of components are measured rather than the change in levels.
Hence the hypothesis tests in this paper are indirect tests of the effects of changes in pension plan components
on AMPL disclosure. Nonetheless, if the two groups of firms exhibit a significant difference in the level of a
key component, then the difference should be consistent with the true effect of a change in level and thus
enable us to determine whether the traditional or component view of AMPL is more appropriate.
The above analysis in which independent variables are examined one at a time for their effect on the
dependent variable has an important limitation. If the independent variables are correlated, then the analysis
may be incomplete without also considering their joint effects. For firms with underfunded plans in our
sample, the correlation between PA and PAPC (both scaled by PBO to remove any size bias) is about 0.8 in
any fiscal year, and the correlation between PA and UGL (both scaled by PBO) is about -0.4 to -0.5 in any
fiscal year. A priori the test statistics for hypotheses 4 and 5 may be positively correlated while the test
statistics for hypotheses 3 and 4 may be negatively correlated. The results should be interpreted with this in
mind. The effect of the correlations of PAPC and UGL with PA is mitigated by the fact that the predicted
effect of PA on AMPL is the same under both the traditional and component views. All other pairs of independent
variables have insignificant correlations.
Sample and Data
This study draws data from the edition of Research Insight (formerly COMPUSTAT) current through
November 2001. Additional minimum pension liability was established by SFAS No. 87, paragraphs 36 to
38. While other provisions of SFAS No. 87 were implemented for “fiscal years beginning after December 15,
1986” (SFAS No. 87, para 76), AMPL was not required to be measured until “fiscal years beginning after
December 15, 1988” (SFAS No. 87, para 76). Thus for firms with calendar year fiscal years, the first year of
recognition of AMPL is 1989, and the first Research Insight fiscal year in which all companies are required to
measure AMPL is fiscal year 1990. This paper begins its analysis with 1993, several years after the original
implementation, to allow companies to adjust to this change in the disclosure environment. In June 1997 the
exposure draft predecessor of SFAS 132, “Employers’ Disclosures about Pensions and Other Postretirement
Benefits,” was issued primarily to coordinate and streamline the disclosures of pensions under SFAS No. 87
and of other post-retirement benefits under SFAS No. 106. In February 1998 the final version of SFAS No.
132 was issued. It introduced detailed reconciliations of projected benefit obligation and plan assets but
eliminated and combined some disclosures related to accumulated benefit obligation.1 This paper therefore
limits its analysis to Research Insight fiscal years 1993–1996 (fiscal years ending June 1993 through May
1997) in order to preserve a common disclosure environment across the sample.
59
Southwest Business and Economics Journal/2003-2004
The population is restricted to independent and publicly traded operating firms incorporated in the
United States. A firm is included in the sample for a given fiscal year if data for the fiscal year meets the
following conditions (Research Insight data item in parentheses). Four key year end financial variables must
be positive: market capitalization (MKVALF), common shares outstanding (CSHO), total assets (AT), and
total sales (SALE). Four other key financial variables must be present: stockholders’ equity (SEQ), total debt
(DT), income before extraordinary items (IB), and net income (NI). The population excludes closed-end
funds and unit trusts (SIC code 6726), royalty traders and Real Estate Investment Trusts (SIC codes 67906799), and other nonoperating establishments (SIC code 9995). The population also excludes nonindependent
firms (i.e., subsidiaries) identified by CUSIP coding (99 in digits 5 and 6 and 00 in digits 7 and 8 of the
CUSIP number).2
In each fiscal year, the population is subdivided at three levels. The first subdivision is defined by
the firm’s projected benefit obligation (PBO), which is computed as projected benefit obligations for overfunded
plans (PBPRO), plus projected benefit obligations for underfunded plans (PBPRU).3 The population is
subdivided into firms with no defined benefit pension plan (PBO = 0 or missing) versus those with such plans
(PBO > 0). The subpopulation of firms with PBO > 0 is subdivided into those with fully funded plans (generally,
those with PBPRU = 0 or missing) and those with underfunded plans (generally, those with PBPRU > 0).4
Finally, the subpopulation of firms with underfunded plans is subdivided into those with no AMPL (PADDML
= 0 or missing) versus those with AMPL (PADDML > 0). Hypothesis tests in this study concern these last two
subgroups.
Companies with defined benefit pension plans are significantly different from those without such
plans. The former represents a self-selected group of companies that chose to offer their employees a defined
benefit pension plan. Otherwise, they represent the full spectrum of companies in terms of size, profitability,
and capital structure. However, these firms are more likely to be manufacturing, transportation, and
communications companies and less likely to be service companies than firms with no defined benefit pension
plans. See Table 1 for Research Insight fiscal year 1996. Companies with defined benefit pension plans tend
to be larger companies. As a group, they have significantly greater market capitalization, total assets, total
debt, total sales, net income, and number of employees, on average, than firms without defined benefit pension
plans. See Table 2 for Research Insight fiscal year 1996. (Statistics in Tables 1 and 2 for other fiscal years are
similar.)
Companies with underfunded pension plans are significantly different from those with fully funded
plans. Firms with underfunded plans are more likely to be manufacturing companies and less likely to be
transportation, communications, or financial service companies (Table 1). Companies with underfunded benefit
pension plans tend on average to have significantly more employees, greater sales, lower net operating assets
to sales, and higher periodic pension cost (Table 2).
Companies with and without AMPL have roughly similar industry representation (Table 1). However,
firms with AMPL tend to have significantly more employees, greater projected benefit obligation, more
pension plan assets, and higher periodic pension cost, on average, than companies with no AMPL (Table 2).
Contrary to some assertions in the literature, AMPL has been neither rare nor negligible. For Research
Insight fiscal year 1996, 368 firms disclosed AMPL. The interquartile range for AMPL was $0.676 million–
$9.925 million, and the median value was $2.463 million. Statistics for the other fiscal years in this study are
similar.
Most pension fund variables in this study are computed by adding the corresponding Research
Insight items for overfunded and underfunded plans at each firm:
•
UPSC = unrecognized prior service cost = PCUPSO + PCUPSU;
•
UGL = unrecognized net losses (gains) = POAJO + POAJU;
•
PA = pension plan assets at fair value = PPLAO + PPLAU;
•
PAPC = prepaid (accrued) pension cost = PCPPAO + PCPPAU;
•
ABO = accumulated benefit obligation = PBACO + PBACU.
In addition, one other pension plan item was collected:
60
Additional Minimum Pension
Liability and Pension Underfunding
•
RCI = estimated rate of compensation increase in percent (PPRCI in Research
Insight).
The pension fund variables in this study are aggregated across all defined benefit pension plans at a
firm. Although pension plans are restricted funds, this study assumes that companies manage their pensions
as a whole. Thus AMPL is a company liability rather than simply an underfunded measure of a particular
pension plan. In fact, information about the specific pension plan that gives rise to AMPL need not be disclosed.
Although pensions are accounted for on a plan-by-plan basis, disclosures are for pension plans grouped by
their funding status. So even the information disclosed about underfunded pension plans is not necessarily
related directly to the amount of AMPL recorded.
For Research Insight fiscal year 1996, the projected benefit obligations for firms with defined benefit
pension plans (PBO > 0) range from $0.293 million to $82,027 million. The range for other fiscal years is
similarly wide. Thus, with the exception of RCI, comparison of the absolute level of variables is inappropriate
because a size effect may bias the results. Pension variables in the hypothesis tests are scaled by PBO because
it represents the best theoretical measure of pension plan obligation. Additional tests conducted with number
of employees, accumulated benefit obligation, and total assets as the scaling variables produced results similar
to those reported in this paper.
Missing data items in Research Insight plague the study of defined benefit pension plans. When a
company discloses only overfunded plans or only underfunded plans, Research Insight rarely makes the
logical deduction that no funds of the other variety exist. In situations when a researcher would record zero as
values for the overfunded (or underfunded) versions of UPSC, UGL, PA, PAPC, ABO, and PBO, Research
Insight typically lists the missing data code (@NA). Therefore, for each of these variables in this study,
companies with complete data for the overfunded (underfunded) pension plan item and the missing data code
for the underfunded (overfunded) item are assumed to have zero values for the missing item.
In addition, Research Insight assigns the missing value code to RCI for some companies with otherwise
complete data. This situation may arise when a company reports a range of values or separate values for
different types of defined benefit pension plans at the firm. An informal sampling of actual annual reports
underlying the Research Insight database reveals uncaptured pension footnote data of this sort. To address
this problem, we repeat the test of Hypothesis 1 with missing values of RCI imputed as 100(PBO/ABO - 1)
provided that the imputed value is positive.5 An alternative approach that we did not take to address this
missing data problem would be to manually collect pension footnote data and then compute the midpoint of
ranges reported for RCI rather than omitting this information or imputing it (the two approaches in the current
study).
Results
The cross-sectional hypothesis tests tend to support the component formulation of AMPL and generally
are inconsistent with the traditional formulation. For at least two of the four fiscal years, the null hypothesis
in Hypotheses 1, 2, and 3 (tests of RCI, UPSC, and UGL, respectively) is rejected in favor of the alternative
that is consistent with the component formulation. The evidence is strongest in the tests on UGL, where the pvalues are smaller than 0.001 (see Table 3). Null Hypothesis 5 (the test of PAPC) is not rejected at the 20%
level in three of the four fiscal years, except 1996, when it is rejected in favor of the alternative that is
consistent with the traditional formulation. Of the 16 hypothesis tests (H1, H2, H3, and H5 for four fiscal
years each), only one supports the traditional over the component formulation while eight support the component
over the traditional formulation.
Rate of Compensation Increase (RCI) has a greater frequency of missing values than the variables in
the other hypothesis tests. About one third of all firm/fiscal year observations in the underfunded plan sample
are missing RCI. To check whether a missing data bias exists, we repeated the test of Hypothesis 1 with each
missing value of RCI imputed as 100(PBO/ABO - 1), provided this value was positive. The null hypothesis
is rejected in all four years in favor of the alternative that is consistent with the component formulation.
61
Southwest Business and Economics Journal/2003-2004
The results in Table 3 are relatively robust. Alternative scaling factors—number of employees,
accumulated benefit obligation, or total assets—yield similar results to those from tests where PBO is the
scaling factor. Nonparametric tests of the hypotheses using the adjusted Wilcoxon test (also known as the
Mann-Whitney test) generally confirm the results of the parametric tests reported in Table 3.
Conclusion
In summary, the traditional view of AMPL as a function of accumulated benefit obligation, plan
assets, and prepaid (accrued) pension cost has little support in this study as a useful empirical explanation of
AMPL. Instead, AMPL is more directly related to the underlying components of unrecognized prior service
cost and unrecognized net losses (gains). This decomposition of AMPL suggests that the meaning of AMPL
to a firm may depend on which of these two components is more responsible for the AMPL status. Companies
can choose to increase unrecognized prior service cost through negotiations with employees; companies can
adjust or fail to adjust to external changes in their pension plans to create unrecognized net losses (gains).
Additional minimum pension liability may mean different things to different decision-makers, depending on
the interplay of these two factors. Future studies will look at the individual firm time series of unrecognized
prior service cost, unrecognized net losses (gains), and AMPL status to identify the changes that prompted
recognition of AMPL.
End Notes
1
See Luecke and Andrzejewski (1998) for a summary of the SFAS No. 132 disclosure changes.
The population also excludes 36 firms identified by examining company names for multiple issues, proforma duplicate listings, and nontrading consolidated statements.
3
If only one of the two variables is available, we impute the other as zero. If both are missing, we infer that
the firm had no defined benefit pension plans in that fiscal year.
4
For most firms with underfunded plans in the period covered by the analysis in this paper, Research Insight
reports values for PBPRU and other pension fund data items associated with underfunded plans at the firm.
Hence, PBPRU > 0 usually flags a firm with underfunded
plans at the firm. Hence, PBPRU > 0 usually flags a firm with underfunded plans. However, for a few firms
(15 to 18 per year) in our sample, Research Insight apparently combines overfunded and underfunded amounts
in PBPRO and other pension fund data items usually associated with overfunded plans but also reports a
positive value for PADDML, Research Insight’s data item for AMPL. To assure that these firms are correctly
categorized, we assign firms with PBO > 0 to the subpopulation with underfunded plans if one of two conditions
exist: (a) if PBPRU > 0; or (b) if PBPRU is missing but PADDML > 0. Otherwise a firm with PBO > 0 is
classified as having fully funded plans.
5
For Research Insight fiscal years 1993-1996, about 37% of the underfunded firm/fiscal year observations are
missing RCI. The other variables had missing data rates of less than 1%.
2
62
Additional Minimum Pension
Liability and Pension Underfunding
Table 1. Industry Composition of Population.
Firms are publicly traded U.S. companies.
Fiscal years ending June 1996–May 1997 (Research Insight fiscal year 1996).
Economic Sector No Defined Benefit Defined Benefit
by SIC Code
Pension Plans
Pension Plans
No.
%
Fully Funded
Plans
Under -funded
Plans
No AMPL
With AMPL
No.
%
No.
%
No.
%
No.
%
No.
%
Agriculture,
forestry, fishing
14
0.3
5
0.4
1
0.1
4
0.6
2
0.7
2
0.5
Mining &
Construction
224
5.0
56
4.0
29
4.2
26
3.9
11
3.8
15
3.9
12.4
288
20.7
116
16.7
166
24.7
56
19.5
110
8.6
26.0
455
32.7
176
25.4
270
40.2
115
40.1
155 40.4
15.1
158
22.8
50
7.5
24
8.4
26
6.8
Manufacturing:
556
consumercyclicals
Manufacturing:
1,165
consumer staples
Transportation,
communications
265
5.9
210
Wholesale
& retail trade
526
11.7
120
8.6
58
8.4
61
9.1
28
9.8
33
8.6
Finance,
760
insurance, realestate
Services: hotels, 734
personal, business
17.0
181
13.0
122
17.6
55
8.2
29
10.1
26
6.8
55
4.0
23
3.3
31
4.6
17
5.9
14
3.6
Services: health, 238
legal, social, etc.
5.3
7
1.0
6
0.9
5
1.7
1
0.3
4
0.6
2
0.3
0
0.0
2
0.5
100.0
671
287
100.0
16.4
Public
administration
0
0.0
TOTAL
4,482
100.0
14
6
1,390
1.0
0.4
100.0
694
100.0
384 100.0
Notes. SIC codes are approximate for these fiscal years. Research Insight reports current SIC codes but not historical codes. 2001
was the current year for the database from which this sample was drawn. Codes: agriculture, forestry, fishing: 0001-0999; mining
& construction: 1000-1999; manufacturing of consumer-cyclicals: 2000-2999; manufacturing of consumer staples: 3000-3999;
transportation, communications: 4000-4999; wholesale & retail trade: 5000-5999; finance, insurance, real estate: 6000-6999;
hotels, personal, business service: 7000-7999; health, legal, social services: 8000-8999; public administration: 9000-9999. The
subsample of companies with defined benefit pension plans is subdivided into those with fully funded plans and those with
underfunded plans. The subsample of companies with underfunded plans is subdivided into those with no additional minimum
pension liability (AMPL) and those with AMPL. Fiscal years 1993-1995 have similar sample distributions.
63
Southwest Business and Economics Journal/2003-2004
Table 2. Financial Characteristics of Population.
Firms are publicly traded U.S. companies.
Fiscal years ending June 1996–May 1997 (Research Insight fiscal year 1996).
Median values of financial variables and t-tests for differences in the means.
Employees in 1000s; all other variables except ratios are in $ millions.
Employees
No
Defined
Benefit
Plans
Median
0.266
Defined
Benefit
Plans
Median
3.601
Financial Statement Items
Market
58.528
583.352
Capitalization
t-stat for
difference
in means
-11.06
Fully
Funded
Plans
Median
2.765
Underfunded
Plans
t -stat for
difference
in means
No
AMPL
With
AMPL
Median
5.697
t -stat for
difference
in means
Median
4.830
-2.99
Median
3.900
-1.97
528.080
717.848
-0.57
0.94
653.638
911.833
-0.71
-10.84
549.961
633.343
784.4
-2.74
Total
Assets
59.132
802.243
-9.37
803.931
Total
Debt
4.350
184.056
-5.68
187.344
179.184
0.75
149.600
221.056
-0.21
242.400
255.729
-0.75
Stockho26.802
lders’ Equity
272.197
-13.76
280.230
249.937
-0.44
Total
Sales
40.146
696.262
-11.51
598.066
805.374
-2.48
649.300
985.155
-1.52
Net
Income
1.044
30.833
-11.35
30.916
29.912
-1.71
27.335
31.063
0.00
0.589
2.84
0.678
0.527
2.95
0.563
0.513
1.50
Financial Ratios
Net Oper0.539
ating Assets
to Sales
Debt Ratio
0.130
0.255
-3.68
0.256
0.253
-2.45
0.246
0.266
-1.85
Return on
Assets
0.014
0.041
-10.06
0.039
0.046
0.38
0.046
0.046
-0.10
Pension Plan Items
Projected Benefit
Obligation
53.871
70.787
-0.99
43.061
102.124
-2.46
Pension Plan Assets
55.222
63.566
-0.32
38.679
86.798
-2.64
Periodic Pension Cost
0.685
1.822
-3.91
1.434
2.329
-1.89
694
671
287
384
Sample Size 4,482
1,390
Notes. Net operating assets to sales ratio calculated as (stockholders’ equity + total debt - cash)/(total sales), where Research Insight
CHE was used for cash. Calculations using cash plus investments and advances (CHE plus IVAEQ variables in Research Insight)
yield similar results. t-statistics computed assuming unequal variances. The sample size for some variables was smaller than the
value above. In particular, a small number of firms in the above sample were missing one or more of the following variables:
employees, net operating assets to sales ratio, pension assets, and periodic pension cost.
64
Additional Minimum Pension
Liability and Pension Underfunding
Table 3. Results of Cross-sectional Hypothesis Tests.
Firms are publicly traded U.S. companies.
Fiscal years ending June 1993–May 1997 (Research Insight fiscal years 1993–1996).
Hypotheses
0: subpopulation with
AMPL = 0;
+: subpopulation with
AMPL > 0
Fiscal
Year
1993
H10: mean RCI(0) <
mean RCI(+)
H1A: mean RCI(0) >
mean RCI(+)
1994
1995
1996
1993
H20: mean UPSC(0)>
mean UPSC(+)
H2A: mean UPSC(0) <
mean UPSC(+)
1994
1995
1996
1993
H30: mean UGL(0)>
mean UGL(+)
H3A: mean UGL(0) <
mean UGL(+)
1994
1995
1996
Mean of Test Statistic
(Standard Deviation) [Sample
Size ] for
Subpopulation
AMPL=0
AMPL>0
t-statistic
(1-sided
p-value)
4.8871
(1.0169)[149]
4.9908
(0.9439)[169]
4.7795
(0.9274)[176]
4.6671
(0.9664)[175]
4.8601
(1.1481)[259]
4.8165
(1.1784)[246]
4.6220
(1.0974)[273]
4.5662
(1.0867)[239]
0.24
(0.406)
1.60
(0.055)
1.58
(0.058)
0.98
(0.164)
0.0260
(0.0579)[229]
0.0295
(0.0791)[264]
0.0324
(0.0970)[279]
0.0292
(0.0801)[285]
0.0403
(0.0683)[393]
0.0449
(0.0820)[371]
0.0369
(0.0724)[430]
0.0359
(0.0776)[381]
-2.66
(0.004)
-2.37
(0.009)
-0.71
(0.239)
-1.09
(0.138)
0.0503
(0.1549)[229]
0.0103
(0.1575)[265]
0.0235
(0.1566)[279]
-0.0173
(0.1632)[286]
0.0964
(0.1634)[394]
0.0781
(0.1666)[ 371]
0.0893
(0.1641)[431]
0.0379
(0.1770)[381}
-3.45
(0.0003)
-5.18
(0.0000)
-5.31
(0.0000)
-4.11
(0.0000)
Consistent with Formulation
of AMPL:
Traditional
Component
n.i.
n.i.
No
Yes
No
Yes
n.i.
n.i.
No
Yes
No
Yes
n.i.
n.i.
n.i.
n.i.
No
Yes
No
Yes
No
Yes
No
Yes
65
Southwest Business and Economics Journal/2003-2004
Table 4. Results of Cross-sectional Hypothesis Tests.
Firms are publicly traded U.S. companies.
Fiscal years ending June 1993–May 1997 (Research Insight fiscal years 1993–1996).
Hypotheses;
0: subpopulation with
AMPL = 0
Fiscal
Year
+: subpopulation with
AMPL > 0
1993
H40: mean PA(0)< mean
PA(+)
H4A: mean PA(0) > mean
PA(+)
1994
1995
1996
1993
H50: mean PAPC(0)>
mean PAPC(+)
H5A: mean PAPC(0) <
PA(+)
mean PAPC(+)
1994
1995
1996
Mean of Test Statistic
(Standard Deviation) [Sample
Size] for Subpopulation
t-statistic
(1-sided
p-value)
AMPL=0
AMPL>0
0.8004
(0.2814)[229]
0.8431
(0.2647)[266]
0.8227
(0.2852)[281]
0.8389
(0.3119)[287]
0.8097
(0.2425)[395]
0.8162
(0.2472)[372]
0.8141
(0.2513)[432]
0.8872
(0.2686)[284]
-0.43
(0.668)
1.32
(0.094)
0.42
(0.336)
-2.15
(0.984)
-0.1441
(0.2003)[395]
-0.1461
(0.2137)[372]
-0.1447
(0.2157)[432]
-0.1111
(0.2152)[384]
1.45
(0.874)
1.62
(0.947)
1.36
(0.913)
-2.02
(0.022)
-0.1233
(0.2475)[229]
-0.1163
(0.2475)[266]
-0.1209
(0.2464)[281]
-0.1487
(0.2659)[287]
Consistent with Formulation of
AMPL:
Traditional
Component
n.i.
n.i.
Yes
Yes
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
Yes
No
Notes. RCI = estimated rate of compensation increase in %. UPSC = unrecognized prior service cost. UGL = unrecognized net losses
(gains). PA = pension assets. PAPC = prepaid (accrued) costs. Test results are reported for UPSC, UGL, PA, and PAPC scaled by
projected benefit obligation (PBO). t-statistics for UPSC, UGL, PA, and PAPC scaled by number of employees or by total assets are
similar to the above. t-statistic and one-sided p-value are computed assuming equal variance. (Results assuming unequal variances
are very similar.) If a one-sided t-test rejects the null hypothesis in favor of the alternative at the 10% level or better, then we write
“Yes” if the alternative is consistent with the formulation and “No” if it is not. If a one-sided t-test does not reject the null hypothesis,
then, strictly speaking, we have no formal statistical evidence in favor or against either formulation. So we write n.i. (not inconsistent
with the indicated formulation of AMPL).
66
Additional Minimum Pension
Liability and Pension Underfunding
References:
Amir, E. & Benartzi, S. (1999 Summer). Accounting Recognition and the Determinants of Pension Asset
Allocation. Journal Of Accounting, Auditing & Finance, 321-343.
Bader, L. N. (2003 May/June). Treatment of Pension Plans in a Corporate Valuation. Financial Analysts
Journal, 19-25.
Blitzer, D. M., Friedman, R. E. & Silverblatt, H. J. (2000 May 14). Measures of Corporate Earnings. New
York: Standard & Poor’s.
Bryan-Low, C. (2002 November 26). Pension Liability Raises a Flag Among Bankers. The Wall Street
Journal, C1.
Byrnes, N. (2000 April 24). The Perils of Fat Pension Plans Investors Beware: Corporate Income is Being
Puffed Up. Business Week, 91-92.
Chen, Y. H. & Lin, W. T. (1992). Pension Benefits Effects on Capital, Market Equilibrium, Firm Value,
and Financing Decisions. Journal of Business Research, 1-25.
Coronado, J. L. & Sharpe, S. A. (2003). Did Pension Plan Accounting Contribute to a Stock Market
Bubble? Brookings Panel on Economic Activity, Volume 1.
Dale, A. (2002 December 11). Audit Watchdog Targets Pensions. The Wall Street Journal, p. B5.
Financial Accounting Standards Board. (1995). Disclosure Effectiveness. Stamford Connecticut: Financial
Accounting Standards Board.
Financial Accounting Standards Board. (1997). Exposure Draft: Employers’ Disclosures about Pensions
and Other Postretirement Benefits. Stamford Connecticut: Financial Accounting Standards Board.
Financial Accounting Standards Board. (1985). SFAS No. 87: Employers’ Accounting for Pensions.
Stamford Connecticut: Financial Accounting Standards Board.
Financial Accounting Standards Board.(1998). SFAS No. 132: Employers’ Disclosures about Pensions and
Other Postretirement Benefits — an amendment of FASB Statements No. 87, 88, and 106.
Stamford Connecticut: Financial Accounting Standards Board.
Hand, J. R. M., Landsman, W. R. & Monahan, S. J. (1998). What Can SFAS No. 87 Tell Us About
Disclosure Effectiveness? Working Paper, University of North Carolina.
Harper, R. M. Jr. & Strawser, J. R. (1993 November). The Effect of SFAS 87 on Corporate Funding of
Defined Benefit Pension Plans. Journal of Business, Finance, and Accounting, 815-833.
Henry, D. (2001 August 13). Why Earnings Are Too Rosy. Business Week, 68-69.
Kemp, R. S. (1987). An Examination of the Relationship of Unfunded Vested Pension Liabilities and
Selected Elements of Firm Value. Advances in Accounting, 59-71.
Kranhold, K. & Schultz, E. (2003 February 21). Leading the News: GE to Disregard Pension Income for
Top Salaries. The Wall Street Journal, p. A3.
Luecke, R. W. & Andrzejewski, C. (1998). FASB 132: What Companies Must Disclose. Journal of
Accountancy, 43-49.
MacDonald, E. (2003 June). Pension Pangs. Forbes, 140-142.
McGough, R. & Schultz, E. E. (1999 September). How Pension Surpluses Lift Profits. The Wall Street
Journal, p. C1.
Newell, G. E., Kreuze, J. G. & Hurtt, D. (2002 Fall). Corporate Pension Plans: How Consistent Are the
Assumptions in Determining Pension Funding Status? Mid-American Journal Of Business, 23-30.
Sharad, A. (1999 Spring). Determinants of Funding Strategies and Actuarial Choices of Defined-Benefit
Pension Plans. Contemporary Accounting Research, 39-74.
Tunick, B. (2003 October 14). The Looming Pension Liability: Analysts focus on problem that FASB
Rules Ignore. The Investment Dealers’ Digest: IDD, 9-10.
67